Overview
- Two BrainGate trial participants with paralysis used an intracortical implant to type from neural signals, with results published in Nature Neuroscience.
- One participant reached 110 characters per minute, about 22 words per minute, with a word error rate near 1.6%, while the other achieved 47 characters per minute.
- The system maps motor-cortex activity for up, down, or curled finger positions to QWERTY keys and refines output with a predictive language model.
- Calibration required roughly 30 sentences, and researchers reported successful sessions in participants’ homes, indicating potential for at-home use.
- Performance varied with diagnosis and electrode placement, and the approach remains early-stage, invasive, and unapproved for broad clinical use as the trial continues.